The CIA director ranks advanced AI on par with nuclear weapons.

Speaking about national security, CIA Director John Ratcliffe likened high-performance AI models to nuclear weapons. Some of the latest high-performance AIs are adept at finding vulnerabilities in existing systems, threatening core national systems such as finance and military systems.

CIA boss compares cutting-edge AI to nuclear weapons
https://www.france24.com/en/live-news/20260630-cia-boss-compares-cutting-edge-ai-to-nuclear-weapons

Speaking at the Amazon Web Services Summit , a cloud innovation information-sharing summit held in Washington, D.C., Secretary of State Ratcliffe, while discussing U.S. cybersecurity, pointed out that 'comparing the capabilities of AI to nuclear weapons would not be inaccurate.' Secretary Ratcliffe also said that 'in discussions with advisors for national security and economic security, we are talking about the impact of cutting-edge AI models,' sharing information that the government is strengthening its security against AI.

In recent years, AI models with superior ability to detect system vulnerabilities, such as 'Claude Mythos,' have emerged, raising concerns that existing systems could be threatened and necessitating countermeasures. Claude Mythos was initially restricted by the US government after its public release, but this restriction was later lifted.

US government lifts restrictions on Claude Fable 5 and Claude Mythos 5 - GIGAZINE

U.S. President Donald Trump has indicated a policy of encouraging AI innovation within the government while also preparing the government to assess the safety of AI, and is working to lay the groundwork for sharing cutting-edge 'frontier models' developed by AI companies with various government agencies.

President Trump signs executive order to strengthen cyber defense capabilities while rejecting AI regulations - GIGAZINE

Secretary Ratcliffe also stated that 'since taking office, monitoring emerging technologies, along with tracking China's activities, has been one of my top priorities,' and accused adversaries of trying to steal American technology.

China, an adversary of the United States, has developed AI models that are superior to those of American companies in their ability to detect vulnerabilities.

China's open-weight model 'GLM-5.2' surpasses Claude Code in vulnerability detection benchmarks - GIGAZINE